Immune checkpoint molecules B7-H6 and PD-L1 co-pattern the tumor inflammatory microenvironment in human breast cancer

B7-H6 and PD-L1 belong to the B7 family co-stimulatory molecules fine-tuning the immune response. The present work investigates the clinical effect of B7-H6 protein expression with PD-L1 status and the infiltration of natural killer cells as potential biomarkers in breast tumor inflammatory microenvironment. The expression levels of B7-H6 protein by cancer cells and immune infiltrating cells in human breast cancer tissues and evaluate their associations with PD-L1 expression, NK cell status, clinical pathological features and prognosis were explored. The immunohistochemistry labeling method was used to assess B7-H6 and PD-L1 proteins expression by cancer and immune cells. The associations between immune checkpoint, major clinical pathological variables and survival rates were analyzed. B7-H6 protein was depicted in both breast and immune cells. Results showed that Tumor B7-H6 expression is highly associated with Her-2 over expression. B7-H6 + immune cells are highly related to the Scarff–Bloom–Richardson grade and associated with PD-L1 expression and NK cells status. Survival analysis revealed a better prognosis in patients with low expression of B7-H6 by cancer cells. Conversely, B7-H6 + immune cells were significantly associated with longer survival. Findings strongly suggest an interaction between B7 molecules that contributes to a particular design of the inflammatory microenvironment. This may influence the efficiency of therapies based on antibodies blocking the PD-L1/PD1 pathway and can explain the detection of clinical benefits only in a fraction of patients treated with immune checkpoint inhibitors.


Results
B7-H6 and PD-L1 immunodetection in breast cancer. The immunohistochemistry analysis showed positive staining for B7-H6 (Fig. 1a) and PD-L1 (Fig. 1b) both on the membrane and in the cytoplasm of cancer cells, but also in TILs (Fig. 1a,b). Representative examples of B7-H6 and PD-L1immunostaining in breast cancer cells (BCC) and TILs are shown in Fig. 1c,d respectively. B7-H6 and PD-L1 expressions were evaluated successfully by IHC in 156 BC tissues and positive controls at different sites and intensity (Fig. 1c). Based on the area and intensity of cyto-membranous immunostaining in BCC, H-scores were generated as described in methods. In order to investigate the correlation between clinical parameters and the B7-H6 and PD-L1protein expression levels in BCC, we categorized the 156 patients into two major subgroups according to the intensity of B7-H6 and PD-L1immunohistochemical staining. Data showed that B7-H6 BCC expression levels were low in 61% (0 ≤ H-score < 100) and high in 39% of cases (H-score ≥ 100). For PD-L1BCC distribution in our cohort, our results show that PD-L1BCC expression levels were low in 80% (0 ≤ H-score < 100) and high in 20% of cases (H-score ≥ 100) ( Table 1). Regarding TILs density distributions, TILs-B7-H6 + were predominantly localized in the stroma surrounding BCC clusters and abundant in 14% of cases. Conversely, TILs-PD-L1 + were abundant in 60% of cases ( Table 1).

Correlation of B7-H6 and PD-L1 expressions with clinical pathological features. The correla-
tions between clinical parameters and B7-H6/PD-L1 expression in patients are reported in Table 1. An overexpression of B7-H6 by cancer cells in Her-2 positive molecular groups was noticed. Statistical analyses indicate a significant correlation of B7-H6 BCC expression with only Her-2 expression (p < 0.01) and molecular subtypes (p < 0.01). However, we did not find any correlation with SBR grade, tumor size, lymph node invasion or metastasis. For correlation of TILs-B7-H6 status with clinical pathological parameters, we found a significant relation with lymph node status (p < 0.05), ER status (p < 0.01) and SBR grading (p < 0.001). Regarding PD-L1expression, our results display a significant correlation of PD-L1BCC expression with histological type (p < 0.05) and SBR Prognostic significance of B7-H6 expression. In order to look for the prognostic value of B7-H6 expression in women breast cancer, the log-rank survival analyses were performed according to B7-H6 expression levels in BCC or TILs scores after classification and collection of survival data. As shown in Fig. 3a, patients with low B7-H6 BCC expression had a markedly longer OS compared to patients with high B7-H6 expression, however, DFS curves are inverted but almost confused (Fig. 3b). In regards to TILs-B7-H6 status in survival analysis (Fig. 3c,d), both OS and DFS rates of the subgroup with higher TILs-B7-H6 + are longer than those of the subgroup with lower TILs-B7-H6 + , Kaplan-Meier survival curves for OS show significant differences between the two groups (p = 0.017) but no significance with DFS curves (p = 0.128). We then addressed the prognostic relevance of B7-H6 expression by tumor cells or TILs, we investigated survival data among combined groups; group 1 = TILs-B7-H6 Low /B7-H6 BCC Low , group 2 = TILs-B7-H6 Low /B7-H6 BCC High and group 3 = TILs-B7-H6 High /B7-H6 BCC Low/High . As shown in Fig. 3 patients from the third group had better survival rates compared to patients belonging the others two groups with significantly longer OS (Fig. 3e, p = 0.028) but insignificant longer DFS (Fig. 3f, p = 0.166).
Prognostic significance of the concomitant expression of B7-H6 and the others immune inflammatory biomarkers. To better understand the impact of B7-H6 expression in relation with PD-L1 and NK cell status on patient's survival, different subgroups were created taking into account the classifications according to PD-L1 and NK-cell statues. Figure 3g-i show survival curves (OS only) of subgroups generated by combinations between B7-H6 BCC status with PD-L1 BCC or TILs-PD-L1or NK-TILs status. Figure 3j-l show survival curves (OS) of subgroups generated by combinations between TILs-B7-H6 status and PD-L1 BCC or TILs-PD-L1 or NK-TILs status. Regardless of the three types of combinations, patients with B7-H6 BCC High / PD-L1 BCC Low (Fig. 3g) or B7-H6 BCC High /TILs-PD-L1 Low (Fig. 3h) or B7-H6 BCC High /NK-TILs Low (Fig. 3i) expression status have obviously the shorter OS among the others subgroups. Conversely, OS curves of TILs-B7-H6 combination groups show that a longer survival is always observed in the subgroup with the high profile of TILs-B7-H6 associated with high status of either PD-L1 BCC (Fig. 3j) or TILs-PD-L1 (Fig. 3k) or NK-TILs (Fig. 3l). The differences are important in the case of overall survival to achieve significant variations in the combinations TILs-B7-H6 High /NK-TILs High (p = 0.017). Nevertheless, TILs-B7-H6 High group in combination with PD-L1BCC or TILs-PD-L1 groups always shows a better survival (p = 0.058). In regards to multivariate analyses, cox-regression showed a significant effect of TILs-B7-H6 (hazard ratio [HR] = 0.038, 95% confidence interval (CI) = 0.01-0.730, p = 0.018), where all 20 patients with high expression have survived. Furthermore, association of NK-TILs status with OS is also significant in multivariate analyses (Cox-regression) after adjusting for Her2 status (HR = 0.169, CI = 0.039-0.728, p = 0.017).
Prognostic significance of B7-H6 besides others immune inflammatory biomarkers in Her2 positive cancer samples. To better understand the impact of B7-H6 and others immune checkpoint molecules within Her2 positive cancer subtype, we sought correlations between clinical parameters and B7-H6 or PD-L1 expressions by tumor cells and TILs (Table 2), statistical analysis show only significant associations of B7-H6 or PD-L1 BCC expression with SBR grade (p = 0.025 and p = 0.035 respectively) but a limited significant association between TILs-B7-H6 status and lymph node invasion (p = 0.055). Furthermore associations between immunological parameters were performed ( Supplementary Fig. 1), results show the same trends across the whole cohort with the exception of the relationship between B7-H6 BCC expression and TILs-B7-H6 status where boxplots show a limited significant association between them ( Supplementary Fig. 1e, p = 0.051). Similarly, but with less intensity in p-value, data indicate significant correlations between PD-L1 BCC expression and the status of TILs ( Supplementary Fig. 1a www.nature.com/scientificreports/ for patients with high B7-H6 BCC expression (supplementary Fig. 2a-c). In contrast, no difference in OS curves was observed in high or low TILs-PD-L1 presence but patients with high PD-L1 BCC expression tended to have a better survival (supplementary Fig. 2d-f).

Discussion
It remains challenging to build knowledge about the dynamism of the inflammatory tumor microenvironment and to understand the behavior of the main intersecting mediators. A growing number of studies have shown that inflammation plays a critical role in tumorigenesis 23 . The inflammatory tumors microenvironment is characterized by the presence of host leukocytes both in the supporting stroma and in tumor areas 24 . These leukocytes are commonly called Tumor-infiltrating lymphocytes (TILs) which are essential for establishing an immune antitumor response but they may contribute to cancer growth and immunosuppression associated with malignancy 25,26 . TILs can be adaptive immune cells or innate immune cells such as cytotoxic cells particularly NK cells which we are interested in studying in our cancer research field. Moreover, it is clear that the inflammatory status involves several molecules such as immune checkpoints. These molecules participate mutually but differently in cellular response mediated by NK cells, but the profile of these molecules is not well understood. Members of the B7 family have been shown to be important participants in the TME designing notably B7-H1 (or PD-L1) molecule which was widely studied in breast cancer and others solid tumors 27 . Others are less studied such as B7-H6 molecule which appears to be of considerable importance. In this study, we have described the expression of B7-H6 and PD-L1both on cancer cell and TILs in women breast carcinoma. Our results show for the first time the expression of B7-H6 by TILs suitably within solid tumors. We have analyzed the widest part of the biopsy because we are convinced that TMAs tissue formats might not provide sufficient representation of the TME notably, immune infiltrating cells. Previous studies have shown only the expression of B7-H6 by cancer cells but no study have reported their expression by immune cell when analyzes are performed on TMA tissues 18,20,28 . So, one must be vigilant in certain technical aspects which can limit or even distort results, this point was raised previously by Sobral-Leite and colleagues 29 .
Our study is the first to investigate the concomitant profile of B7-H6 and PD-L1expression in breast carcinoma tissues by exploring their relationship first, and then their implication in innate immunity through their associations with NK cells status, and finally by looking for their combined prognostic value. Our results have shown that the two immune checkpoint molecules of interest participate differently in breast carcinoma physiopathology despite the strong association between them. Originally, our data have shown that the biomarker B7-H6 has a different clinical significance depending on its expression either by tumor cells or by infiltrating immune cells. In fact, statistical analysis demonstrated that high levels of B7-H6 in tumor cells are strongly correlated with Her2 expression, this result is in agreement with those of Sun and colleagues 21 . However, high B7-H6 + TILs are strongly associated with SBR grade, ER expression and lymph node invasion. Secondly, patients with high B7-H6 tumor expression had worse survival than those with low B7-H6 expression, this result is similar to data from other researchers in both ovarian cancer and breast cancer 18,21 . Nonetheless, patients with high TILs-B7-H6 expression displayed better survival. Regarding PD-L1biomarker expression, our data showed that its expression by tumor cells or immune stromal cells has the same clinical significance. However, we have noted the strong association of PD-L1BCC and TILs-PD-L1with SBR grade. In addition, patient survival curves according to these two biomarker statuses display the same trends with no differences between strong and weak expressions ( Supplementary Fig. 3a-d). These data are consistent with the findings of several studies, while others have shown the opposite. This point has been reviewed by numerous experts who have justified the discrepancy in clinical significance of PD-L1 immune checkpoint among published reports 27 . Discrepancy can be explained by the use of several technical supports by different teams and the lack of a complete standardized platform. On the other hand, the great heterogeneity between studied populations may explain also the observed differences.
We explored the relationship between B7-H6 and PD-L1and their impact on NK cells recruitment. A strong association between the profiles of PD-L1in tumor or immune cells with B7-H6-TILs status was noticed. In addition, both TILs-B7-H6 and TILs-PD-L1 are highly associated with NK-TILs status. These results consolidate the concept of the mutual crosstalk between tumor cells and immune cells involving different checkpoint molecules that control not only T cell activity, but also NK cells function in a synergistic manner. Knowing that NK cells are associated with better disease outcomes even as minority population in TILs ( Supplementary Fig. 4), these cells play an important role in front of the tumor burden. Additionally, survival tests were conducted on different combining groups based on the expression of B7-H6, the prominent biomarker of this study, with the other immunological parameters that had been selected to associate with (PD-L1 and NK-TILs statuses). The strong expression of B7-H6 by cancer cells probably contribute to the bad progression of the disease among all subgroups. Inversely,  www.nature.com/scientificreports/ the high expression of B7-H6 by the immune cells among different subgroups might be related with a recovery process within the immune system involving NK cells and contributing to better disease outcomes. According to NK-TILs density in cancerous tissues, we have noted the best OS belong to the group of combination with high NK-TILs and TILs-B7-H6 together. Thus, the immune checkpoints B7-H6 and PD-L1 are linked together and certainly with other molecules in a way which modulates the inflammatory tumor microenvironment. We speculate a double role which can polarize the immune response either towards an effective antitumor response www.nature.com/scientificreports/ www.nature.com/scientificreports/ or the opposite by maintaining a chronic inflammatory state leading to immunosuppression. Already, different mechanisms damping the function of NK cells have been shown through B7 molecules action. However, little is known about the expression of these molecules that act differently depending on their localization whether expressed on the cellular membrane or sequestered in the cytoplasm or even secreted. These concepts justify the complexity in understanding the mechanisms controlling the tumor inflammatory microenvironment through B7 molecules. Our results suggest a protective role of B7-H6 and PD-L1 molecules when co-expressions are manifested on lymphocytes infiltrating tumors.These cells probably produce interleukins, such as INF-γ, which modulate NK cells function. Indeed, NK cells are considered to be players of innate immune defenses; but more recently, they have been recognized as immunoregulatory cells, secreting pro-inflammatory cytokines and several chemokines 30 . Thus, NK cells may recruit and may be recruited to inflammatory sites where they can co-localize with other immune cells, including professional presenting cells such as dendritic cells. NK cells can cooperate with them and modulate immune response. Besides the immune surveillance role, tumor neovascularization is highly controlled by cytokines in the microenvironment, which may have an immunosuppressive action resulting in an increase in angiogenic factors in tumor tissues. High expression of inflammatory cytokines in the tumor microenvironment seems to be the reason for the low therapeutic efficacy of current antiangiogenic drugs.
To better elucidate the immunomodulatory mechanism of B7 molecules in immune cell behaviors and consequent cytokines balancing, it is important to combine more precise analyses of their location and to conduct analyses allowing biomarkers co-revelation. Therefore, the immunotherapy targeting checkpoint molecules must be revised. The use of combined immune checkpoint inhibitors may result in a reduction of tumor burden and patient benefit in some cases. In other situations, of its inefficacy might be due to the double role of these two molecules. Targeting other molecules from the B7 family can be an important stake that deserves further investigation to decipher their contribution in innate immunity monitoring. Several molecules must be targeted simultaneously according to the patient characteristics. We speculate that a combinatorial therapeutic strategy based on immunotherapy targeting immune checkpoint molecules, antiangiogenic molecules and anti-inflammatory drug in a scalable and personalized way might improve disease outcome.

Materials and methods
Patients and tissues samples. Formalin-fixed, paraffin-embedded tissue samples were collected from 156 patients with primary invasive breast carcinomas who underwent surgical resection at the Department of Gynecology and Obstetrics of the HediChaker University Hospital (Sfax, Tunisia). All procedures performed in this study were in compliance with the ethical standards of the institutional and the national research committee of Habib Bourguiba University Hospital and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. All specimens were analyzed at the Department of Pathology of the Habib Bourguiba University Hospital (Sfax, Tunisia) and all tumor tissues were confirmed as the serous breast cancer by using haematoxylin and eosin (H&E) staining after surgical resection. Clinical-pathological data of patients were retrieved from the hospital's electronic records and available paper records. They included age, histological grade, histological type, molecular subtype, tumor size, lymph node status, distant metastasis, and lymphovascular invasion. All tumors were graded according to Elston Ellis modification of the Scarff-Bloom-Richardson (SBR) histological grading system 31 . The clinical stage was determined according to TNM (tumor, lymph node and metastasis) classification adopted by the International Union Against Cancer 32 . Outcomes data were collected from patient follow-ups at the Department of Medical Oncology of the Habib Bourguiba University Hospital (Sfax, Tunisia). Overall survival (OS) was defined as time in days from the beginning of the study observation period until death. If the death event did not occur, then the OS was noted as the total study observation period in days. Disease free survival (DFS) was defined from the date of the end of treatment until recurrences or last follow-up.  www.nature.com/scientificreports/ Immunohistochemistry. Formalin-fixed, paraffin-embedded tissue samples were cut into 5-μm-thick section. Before immunostaining, haematoxylin and eosin-stained standard slides were reviewed by experimented pathologist (S.C) from each section of breast cancer tissues. A representative tumor region and the corresponding formalin-fixed paraffin-embedded tissue block were selected for use in the tissue array block containing six fragments of 7 mm in diameter.
In order to validate the immunohistochemical staining for special biomarker, striated muscle and placenta tissues were used as positive controls for the revelation of B7-H6 and PD-L1 respectively according to the instructions proposed by the manufacturers. For immunohistochemistry (IHC), each tumor array was cut into 2-μm sections and mounted on Leica Microsystems BOND Plus slides and dried overnight at 60 °C. Immunohistochemistry steps were performed as described in our previous study 33 . Briefly, tissue slides were deparaffinized in xylene followed by subsequent rehydration through graded alcohols then washed in purified water. Heat-induced antigen retrieval was performed at 95 °C for 20 min in in a pH = 9 epitope retrieval solution (Leica Novocastra) for 40 min. After heating, slides were allowed to cool down to room temperature and were briefly washed in phosphate-buffered saline (PBS) solution. Immunohistochemical staining was performed using the Novolink Polymer Detection System (RE7150-K, Leica Biosystems). Briefly, endogenous peroxidase activity was neutralized by Peroxidase Block for 5 min. Slides were washed with PBS, followed by application of Protein Tumor immunostaining scoring and molecular subtyping. The B7-H6 immunostaining densities in tumor cells were analyzed according to the H-score method which has been described in a previous publication 34 .
Briefly, H-score was assessed on the basis of the percentage of positive tumor cells with clearly brown cytoplasm and/or membrane immunostaining. H-score was calculated as follows = (% tumor cells unstained × 0) + (% tumor cells stained weak × 1) + (% tumor cells stained moderate × 2) + (% tumor cells stained strong × 3), and it ranged from 0 (100% negative tumor cells) to 300 (100% strong staining tumor cells). B7-H6 expression was dichotomized into two groups according to the frequency distributions of the H-scores displaying the median as the cut-off value (H-score 0-99 = negative/low expression, and 100-300 = positive/high expression). Frequency and staining intensity of PD-L1 by tumor cells were analyzed, and PD-L1 expression was quantified using the H-score method according to the previous publications 22,35 .
Breast cancer molecular classification is based on the expression of classical biomarkers including estrogen (ER) and progesterone (PR) receptors, the human epidermal growth factor receptor 2 (Her2) and Ki-67 labeling index as a cell proliferation biomarker. Expressions of all biomarkers were carried out using the immunohistochemical method.

Assessment of tumor-infiltrating lymphocytes (TILs).
Based on cellular morphological characteristics, TILs compartmentswere identified on cancerous tissues by haematoxylin staining at the end of the immunohistochemistry protocol. Subsequently, TILs-B7-H6 or TILs-PD-L1 were identified by the corresponding immunostaining through a brown staining on these cells. TILs-PD-L1 were evaluated in at least 3-4 representative high-power field (HPF) areas from each sample, and the mean PD-L1 + TILs count was scored using the following three-tier scoring system: 0-5%/HPF (Low), 5-50%/HPF (Medium), > 50%/HPF (High) 38 . For TILs-B7-H6 assessment, we applied the same strategy deployed in PD-L1enumeration due to the absence of recent publication showing TILs-B7-H6 expression in solid tumor, so we have considered the three-tier scoring described above. TILs CD56 + were assessed differently from TILs-PD-L1and TILs-B7-H6. TILs CD56 + analysis was performed as described in our previous study 33 . Briefly, the number of positive stromal and intratumoral CD56 + lymphocytes was quantified. Scoring of NK-TILs immunostaining was determined as positive or negative by a cut-off value of five cells in ten high power fields (× 40 magnification), yielding an immunoscore (NK-TILs) of Negative (< 5 cells) or Positive (≥ 5 cells).

Statistical analysis.
Statistical analyses were performed using the R language and the SPSS 20.0 statistical software for Windows (SPSS Inc., IBM). Bivariate analysis was performed to assess the correlation between biomarkers and clinical pathological characteristics using Chi-square tests for qualitative variables, Pearson-rank correlation for quantitative variables and ANOVA test for quantitative variation among classes of qualitative variables. The cumulative survival (overall survival, OS; recurrence-free survival, RFS) times were calculated using the Kaplan-Meier method and compared with the log-rank test. A cox-regression was also performed in order to evaluate the significance of prognostic factors on survival in a multivariate context (adjusting for confounding variables). All p-values less than 0.05 were considered significant.